deepseek-harness-brain
Skill9 starsChanged 3 months agoMITView on GitHub
Scaffold and operate DeepSeek Harness Brain, a source-cited Obsidian brain for DeepSeek Harness architecture, operations, extension development, security, testing, and debugging. Use when the user says "deepseek-harness-brain", "DeepSeek Harness Brain", "create a DeepSeek Harness architecture, operations, extension development, security, testing, and debugging brain", "import sources", "synthesize plan", "render report", or wants a persistent vault-backed operating system for DeepSeek Harness architecture, operations, extension development, security, testing, and debugging.
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 8 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 56
Never store credentials in the vault.
- Distinctive: line 57
Never make domain-specific claims without dated trustworthy sources.
- Very specific: line 58
Keep hot, index, overview, and log current.
- Distinctive: line 82
No claim that a keyless, static, build, snapshot, browser, or real-API check proves a stronger tier than it actually ran
- Distinctive: line 83
No invented package behavior, event semantics, defaults, compatibility promise, vulnerability, benchmark, or release status
- Distinctive: line 84
No credentials, tokens, cookies, private settings, or local secret values in brain artifacts
- Distinctive: line 85
No instruction to weaken sandbox, approval, host-trust, or credential boundaries merely to make a command pass
- Distinctive: line 86
No mutation, publication, deployment, account change, or production action from this advisory brain
- Distinctive: line 87
No reliance on archived Agent Notes as current authority when source or current documentation disagrees
- Very specific: line 89
Do not call this brain market-ready unless scripts/audit_brain.py --require market-ready passes and references/release-status.json records …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
DeepSeek Harness Brain
Operate the deployed vault first. Treat CODEX.md, wiki/hot.md, and
wiki/index.md as vault-root-relative paths, where the vault root is the
directory passed to --vault or opened in Obsidian. In this repo, the template
vault root is assets/template-brain/ and the demo vault root is
examples/sample-vault/.
Secretary: use agents/deepseek-harness-secretary.md for grounded answers, claim review,
and vault maintenance. That secretary reads the brain first, cites a vault note
and an official URL, and stays advisory and read-only.
Commands
/deepseek-harness-brain new <client-slug> --owner <name>
/deepseek-harness-brain ingest --vault <path> --file <source>
/deepseek-harness-brain import-checkout --checkout <path> --vault <path> --expected-commit <sha> --as-of <date>
/deepseek-harness-brain synthesize --vault <path>
/deepseek-harness-brain portability --inventory <json> --json-out <json> --markdown-out <md>
/deepseek-harness-brain portability-html --markdown <md> --out <html>
/deepseek-harness-brain report --vault <path>
/deepseek-harness-brain visuals --vault <path>
/deepseek-harness-brain lint --vault <path>
/deepseek-harness-brain hash --vault <path> --check <fixture>
/deepseek-harness-brain next --vault <path>Source checkout equivalent:
deepseek-harness-brain new <client-slug> --owner <name>
deepseek-harness-brain ingest --vault <path> --file <source>
deepseek-harness-brain import-checkout --checkout <path> --vault <path> --expected-commit <sha> --as-of <date>
deepseek-harness-brain synthesize --vault <path>
deepseek-harness-brain report --vault <path> --html-onlyRequired Operating Rules
- Read
<vault>/CODEX.md. - Read
<vault>/wiki/hot.md. - Read
<vault>/wiki/index.md. - Preserve
.raw/as immutable source material. - Never store credentials in the vault.
- Never make domain-specific claims without dated trustworthy sources.
- Keep
hot,index,overview, andlogcurrent. - Record research evidence in
references/source-ledger.json. - Record domain adapter completion in
references/adapter-manifest.json.
Script Mapping
The plugin tool allowlist is intentionally read-only. These commands document bounded repository scripts for the user or an explicitly authorized shell workflow. Loading the skill alone does not grant an agent write authority.
new->python scripts/scaffold_vault.pyingest->python scripts/ingest_source.pyimport-checkout->python scripts/import_deepseek_harness.pysynthesize->python scripts/synthesize_brain.pyportability->python scripts/synthesize_deepseek_harness.pyportability-html->python scripts/render_portability_report.pyreport->python scripts/render_brain_report.pyvisuals->python scripts/generate_vault_visuals.pylint->python scripts/lint_vault.pyhash->python scripts/hash_vault.pynext->python scripts/guide_next_action.py
Quality Gates
- No claim that a keyless, static, build, snapshot, browser, or real-API check proves a stronger tier than it actually ran
- No invented package behavior, event semantics, defaults, compatibility promise, vulnerability, benchmark, or release status
- No credentials, tokens, cookies, private settings, or local secret values in brain artifacts
- No instruction to weaken sandbox, approval, host-trust, or credential boundaries merely to make a command pass
- No mutation, publication, deployment, account change, or production action from this advisory brain
- No reliance on archived Agent Notes as current authority when source or current documentation disagrees
Do not call this brain market-ready unless scripts/audit_brain.py --require
market-ready passes and references/release-status.json records an explicit
human-reviewed market-ready decision. Demo verification is not market readiness.
Community readiness covers safe distribution of this knowledge base. It does not
claim browser, live-provider, native, cross-platform, security-audit, or
production-scale runtime evidence.
Research Refresh
on every DeepSeek Harness release or architecture change; monthly while the project remains in developer preview
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

